AI Search Visibility Agency Dashboard for 15 Clients: Platform for AI Overviews, ChatGPT, and Perplexity Tracking

Fifteen clients is the number where a freelancer spreadsheet dies. You need projects that do not leak prompts, seats that are not one shared login, and a dashboard you can open in a client call without exporting CSV first. The surfaces they care about are usually the same three: Google AI Overviews, ChatGPT, Perplexity. Those three do not live in one official report. Search Console will not store the ChatGPT prompt you typed, and ChatGPT will not file an impression in Search Console, so the agency dashboard is the only place the three surfaces sit next to each other with a shared key.

An "agency dashboard" that only plots mentions will still generate fifteen awkward meetings. Clients ask why they disappeared. You shrug at a line chart, because the line chart does not say whether the model changed, whether the crawler stopped fetching, or whether a competitor published something better. A mention chart with a logo swap is not a dashboard, it is a screenshot with a header.

What 15 clients consume

Each client wants their own prompt list. Mixing a dentist in Utrecht with a B2B SaaS in Austin in one project is how you ship the wrong fan-outs, because the queries that matter to a local clinic and the queries that matter to a software vendor do not overlap and the country targeting does not either. You also want country, and often city, on the prompts, because "best CRM" is not the same answer in two markets and the model knows the difference.

You need engines that match the pitch. ChatGPT and Perplexity on paid Promptwatch plans are in the box. Overviews are too. Otterly's base four include ChatGPT, Overviews, Perplexity, and Copilot; Gemini and Claude cost extra. Profound Starter is ChatGPT-only. If the deck said "we track Overviews and ChatGPT," check the SKU before you sign six retainers on Starter, because the deck and the checkout page can disagree and the client will find out which one was true.

Seats matter more than agencies admit. Passing one password around is not a dashboard, it is a compliance problem and a read-only problem at the same time. Promptwatch agency plans include 10 seats. Brand Essential is 1 seat. That alone decides whether you can put a client marketing manager in as read-only later, and the answer is often the difference between a retainer that scales and one that does not.

White-label and SSO are Enterprise on Promptwatch. Self-serve agency does not invent a branded portal SKU. The workable reporting recipe is Looker Studio via the connector plus per-project keys, not a fake "white label" toggle on Kick-off, and the recipe works because it uses tools the client already trusts rather than a portal you have to defend.

The stack that looks like a dashboard and is not one

Otterly.AI is easy to demo. $29 Lite, 15 prompts, no PDF export. You will screenshot. Data can lag a week. Fine for one SMB. Miserable at fifteen if every extra 100 prompts is $99/mo and Claude is an add-on, because the bill stops being predictable and the freshness stops being defensible in a client call.

Peec AI has agency plans from $245/mo, unlimited seats, and screenshots. Starter at $95 is one project. Fifteen clients is not Starter. Extra models stack on the bill. No crawler logs, no AI conversion tracking, no publish step. We leave it last in ranked roundups, and we leave it there because the category it sits in is monitoring, not action.

Profound looks like the enterprise answer until you price fifteen workspaces and discover the famous coverage is Enterprise-gated. Annual billing on public tiers. A 7-day trial is short for an agency rollout, because fifteen clients do not all hand over their prompt lists in a week.

What we want the dashboard to trigger

This site ranks platforms that act. For an agency, "act" means: see the drop, see whether ChatGPTBot fetched the page, open a gap, draft, get a human to accept, publish if the CMS is Webflow or Framer. Each step in that chain is a thing the dashboard should trigger, not just display, because a dashboard that only displays is a monitor and a monitor does not close the loop.

Promptwatch is the self-serve agency row we use for that. Kick-off is $199/mo with a 7-day trial: unlimited projects and prompts, 10 seats, 10,000 responses, and 10M crawler logs. Growth is $399/mo with 25,000 responses and 25M logs. Scale is $799/mo with 65,000 responses and 100M logs. The shared agency list includes state and city targeting, Data Studio, and custom reports. Shopping insights are not listed as a shared agency entitlement. White-label and SSO still require Enterprise or Custom. Brand plans exist if you only have one site: Essential $95/mo is not an agency product and has no listed crawler-log allowance.

Content Agents include AEO article budgets on brand tiers, at 5, 15, and 30. Do not sell a client "unlimited articles" because Kick-off is unlimited prompts. Those are different meters, and the difference shows up the first month a client asks why only three drafts landed.

Visitor analytics, via script or GTM, is how you stop arguing about "AI traffic" in GA4 with a messy referrer, because the script attributes the visit to the answer rather than to a referrer string the model did not send. Agent Chat is useful in Slack if you connect it; scheduled briefings can skip when nothing changed, which is the right behavior for a tool that lives in a busy channel.

A 15-client rollout that does not pretend

1. One project per client. No shared prompt soup, because a shared prompt list is how one client's queries leak into another client's report.
2. Twenty purchase prompts each, in the client's language, with city if they are local.
3. Connect crawler logs where the CDN allows it, with Cloudflare, CloudFront, Fastly, Vercel, Netlify, and others on the list.
4. Put Looker on Professional+ or agency plans if the client wants a report that is not a live login.
5. Publish only through the review inbox. Fully automated live push exists in the product. We would not turn it on for fifteen retainers, because fifteen retainers means fifteen chances for an unreviewed draft to go live.

Google's AI features documentation is still the Overviews source of truth. GSC generative reports, as of June 2026, list impressions, not a ChatGPT mention log. The agency dashboard is the place those three surfaces finally sit next to each other, with a to-do attached. If your platform cannot attach the to-do, you bought fifteen monitors.